Agent to Agent (A2A) Category Management in FMCG

What is agent to agent (A2A) category management?

Agent-to-agent (A2A) category management is a model where autonomous AI agents on the supplier side and retailer side communicate directly to monitor category performance, exchange insights and coordinate actions in real time. Instead of relying on periodic human meetings and manual reporting, A2A systems continuously analyze shared retail data and surface recommendations within human defined boundaries.

‍

How the A2A protocol works

The A2A protocol is an open standard that enables AI agents built by different vendors to discover each other, delegate tasks and coordinate work across enterprise systems. It was introduced by Google in April 2025, contributed to the Linux Foundation in June 2025, and as of mid-2026 has production deployments at more than 150 organizations.

The protocol defines three foundational elements. Agent Cards, which are machine-readable profiles that describe what an agent can do and what data it can access. Tasks, which are structured units of work that agents exchange. And transport via standard HTTP and JSON, meaning no new protocol layer is required on top of existing infrastructure.

For FMCG category management, this means a supplier agent and a retailer agent built on different platforms, one running on Captain, one running on a retailer's own infrastructure, can communicate and coordinate directly. They do not need to be built by the same vendor. They do not need a human intermediary to translate between systems. The A2A protocol handles the interoperability.

‍

Why AI agents matter in FMCG category management 

Autonomous monitoring: from weekly reports to continuous signals

Today, category teams work on a weekly or monthly reporting cycle. A promotion underperforms for three weeks before it appears in a review. A velocity drop at a specific retailer is invisible until the quarterly category meeting. By then, the competitor has already filled the distribution gap.

A2A category management changes this. The supplier agent monitors all retail accounts continuously. It detects a velocity drop the day it starts. It models the probable cause, whether it is a shelf issue, a competitor promotion or a supply disruption, and surfaces the signal with a proposed action before either the supplier or retailer team has opened their laptop. The human category manager reviews the signal, applies judgment, and approves or adjusts the recommendation.

Cross system execution: from fragmented tools to coordinated action

When a category insight requires action, it currently requires a human to translate that insight into an instruction, send it to the right person, wait for a response, and follow up. Each handoff is a delay. Each delay is a lost opportunity.

In an A2A model, a primary agent can delegate tasks to specialized execution agents across ERP systems, procurement tools, or retailer platforms. A supplier agent that identifies an assortment gap can prepare a proposal, structure it in the retailer's preferred format, and flag it for human approval, all without manual intervention between steps. The category manager approves. The action moves.

Touchless decision making within human defined boundaries

Not every category decision needs human approval at every step. Routine signals, standard reorder recommendations, promotion performance flags within expected parameters, can be handled autonomously within boundaries that the category manager defines upfront.

Having a human in the loop remains essential throughout. The A2A model does not remove human judgment from category management. It removes the manual preparation work that currently consumes that judgment. The category manager sets the boundaries. The agent operates within them. Anything outside those boundaries surfaces for review. Data quality is the foundation of all of this: if the underlying data is wrong, every A2A recommendation built on top of it is wrong too.

‍

Why agent to agent matters specifically for FMCG category management

Most discussions of A2A category management focus on procurement and supply chain: agents that monitor supplier risk, manage purchase orders, or optimize inventory. That is valuable. But it misses the most important application for FMCG suppliers: the category relationship with the retailer.

The supplier-retailer data gap is structural. Retailers own roughly 60% of the relevant data in the value chain. Suppliers own around 20%. As retailers invest more heavily in AI tools, their ability to make category decisions without supplier input grows. If the gap in analytics capability becomes too large, retailers stop relying on supplier input altogether. 

A2A category management is the structural answer to that gap. Not by giving suppliers access to data they do not have, but by ensuring that the data suppliers do have flows continuously, cleanly and in context to the shared layer where category decisions are made. A supplier whose agent contributes reliable, harmonized, near real-time data to a joint A2A system is a partner the retailer cannot afford to exclude. A supplier whose agent contributes fragmented or delayed data is a liability.

‍

What A2A category management requires 

A unified data foundation

A2A category management is only as good as the data the supplier agent works with. That requires a unified data foundation: all retail sources harmonized against a single consistent product model, EAN changes processed automatically so historical trend lines stay intact, and near real-time visibility across all retail accounts. Without that foundation, the supplier agent produces recommendations the retailer agent cannot validate or trust. Data harmonization is the prerequisite for A2A, not an optional upgrade.

Context embedded in the data

The A2A protocol enables agents to share tasks and coordinate autonomously. But the quality of that coordination depends on the context embedded in the data. A velocity drop means something different if there was a shelf reset that week. A promotion spike means something different if a competitor was running a simultaneous discount. That context has to be structured into the data model, not left in a spreadsheet note. The supplier's information manager role, ensuring data is clean, structured and context-rich, is what makes the supplier agent valuable in an A2A system.

Defined data sharing boundaries

A2A communication between supplier and retailer requires clear agreements about what data is shared and what remains internal. The supplier agent has access to commercially sensitive pricing, margin and competitive intelligence. The retailer agent has access to shopper data that retailers treat as proprietary. The shared data layer needs to be explicitly defined before A2A communication begins. Trust is built through transparency about these boundaries, not despite them.

‍

The future role of the category manager in an A2A model 

A2A category management does not eliminate the category manager. It transforms what the category manager does. The commercial relationship manager focuses on the human dimension of the retailer partnership: building trust, negotiating annual deals, making strategic calls that require commercial judgment and context that no agent can supply. 

The information manager ensures that the data the supplier agent works with is clean, structured and rich enough in context to produce reliable recommendations within the A2A system.

The category manager on each side of the A2A system needs clear visibility into what the agents are recommending and why, and the ability to override, adjust or redirect at any point. Agentic systems surface signals and prepare recommendations. The category manager interprets them, applies context, and makes the call. That judgment is not automated away by A2A. It is freed up by it.

‍

How Captain is preparing FMCG suppliers for A2A category management 

Captain is built to give FMCG suppliers the data foundation that A2A category management requires. The platform automatically harmonizes all retail sources against a single consistent product model. EAN changes are processed automatically. Restatements are handled without manual intervention. The supplier agent always has access to data that is current, complete and structured for autonomous use.

The agentic AI layer in Captain already operates within the supplier's data environment: monitoring category performance, surfacing signals, preparing recommendations and answering questions in plain language. The next step is the A2A layer: connecting that supplier agent to the retailer's systems through the open A2A protocol, enabling continuous category collaboration that neither party can achieve alone.

At Elho, automating the data foundation was the first step. Category plans grew from 10 to 25+. At Johma, AI driven promotion analysis led to a plan adopted directly by Hoogvliet, Vomar and Plus. Those results were achieved before A2A. The foundation is already in place. 

‍

Ready to build the foundation for A2A category management?

Request a demo to see how Captain helps your team build the data foundation for agent-to-agent collaboration, and come away with practical tips for your specific situation.